Research Article
Distributed Stochastic Alternating Direction Method of Multipliers for Big Data Classification
@INPROCEEDINGS{10.1007/978-3-030-48513-9_11, author={Huihui Wang and Xinwen Li and Xingguo Chen and Lianyong Qi and Xiaolong Xu}, title={Distributed Stochastic Alternating Direction Method of Multipliers for Big Data Classification}, proceedings={Cloud Computing, Smart Grid and Innovative Frontiers in Telecommunications. 9th EAI International Conference, CloudComp 2019, and 4th EAI International Conference, SmartGIFT 2019, Beijing, China, December 4-5, 2019, and December 21-22, 2019}, proceedings_a={CLOUDCOMP}, year={2020}, month={6}, keywords={Big data ADMM Stochastic ADMM Distributed classification}, doi={10.1007/978-3-030-48513-9_11} }- Huihui Wang
Xinwen Li
Xingguo Chen
Lianyong Qi
Xiaolong Xu
Year: 2020
Distributed Stochastic Alternating Direction Method of Multipliers for Big Data Classification
CLOUDCOMP
Springer
DOI: 10.1007/978-3-030-48513-9_11
Abstract
In recent years, classification with big data sets has become one of the latest research topic in machine learning. Distributed classification have received much attention from industry and academia. Recently, the Alternating Direction Method of Multipliers (ADMM) is a widely-used method to solve learning problems in a distributed manner due to its simplicity and scalability. However, distributed ADMM usually converges slowly and thus suffers from expensive time cost in practice. To overcome this limitation, we propose a novel distributed stochastic ADMM (DS-ADMM) algorithm for big data classification based on the MPI framework. By formulating the original problem as a series of sub-problems through a cluster of multiple computers (nodes). In particular, we exploit a stochastic method for sub-problem optimization in parallel to further improve time efficiency. The experimental results show that our proposed distributed algorithm is suitable to enhance the performance of ADMM, and can be effectively applied for big data classification.
